OK....hmm....

My colleague can't recall where he heard it.  Maybe he made it up, or it was a 
misunderstanding about a recent patch that Alex (author of STAR) issued.  Never 
mind...

In any case, would we need to specify a single large memory node to utilize 
this feature for STAR?  How long does the memory allocation last?

David

--------------------
David Hoover, PhD
Helix Systems Staff
SCB/DCSS/CIT/NIH
301-435-2986
http://helix.nih.gov

On Sep 26, 2014, at 9:21 AM, David Hoover <hoove...@helix.nih.gov> wrote:

> A colleague of mine mentioned it.  I'll ask him where he got his info.  Just 
> to clarify: do you always run STAR jobs on the same host?  We are running 
> Galaxy in front of a batch system cluster, and so by default STAR jobs would 
> run on different nodes.  It's not clear to me how long the memory allocated 
> would last after the batch job finished.  How do you determine whether the 
> memory remains allocated and whether the job has been accelerated due to 
> pre-loaded data?  For example, if you create a genome reference, using 
> --genomeLoad=LoadAndKeep, then run an alignment, are subsequent alignments 
> using the same genome reference much faster?  If so, how much faster?  
> 
> I apologize, I am a jack of all trades, master of none.  I could test this 
> myself, but everything I touch related to genomics takes >50GB of memory and 
> >18 hours clocktime, and it gets painful to try testing everything.
> 
> David
> 
> --------------------
> David Hoover, PhD
> Helix Systems Staff
> SCB/DCSS/CIT/NIH
> 301-435-2986
> http://helix.nih.gov
> 
> 
> 
> 
> On Sep 26, 2014, at 4:13 AM, Ross <ross.laza...@gmail.com> wrote:
> 
>> Hi David.
>> 
>> I've not needed that workflow so haven't a solution for you and no, it 
>> doesn't do anything with chimeric output - won't be hard to add I suspect. 
>> There's no python wrapper - just shell script in the command segment.
>> 
>> It's not in an IUC main tool shed repository because it lacks a data manager 
>> - manual star indexes are a bit of a pain but less pain than writing a data 
>> manager :( so I haven't yet. Might be run best through the API.
>> 
>> On shared memory: Pity. it works a treat for us. I didn't see anything on 
>> the google group - do you recall where you learned about this deprecation ?
>> 
>> 
>> On Thu, Sep 25, 2014 at 10:41 PM, David Hoover <hoove...@helix.nih.gov> 
>> wrote:
>> Ross,
>> 
>> About the index files:  It is way easier to have pre-built index files.  
>> However, when running a 2-pass STAR run, a user will need to generate their 
>> own reference index files based on the output SJ.tab.out file created in the 
>> first pass.  Is this incorporated into your tool?
>> 
>> About shared memory:  I am under the impression that the latest version of 
>> STAR has deprecated this feature.  I am unclear how this would help unless a 
>> single large-memory machine was dedicated to running all STAR jobs.  Is this 
>> the case?
>> 
>> Also, does the tool merge the SAM/BAM file with the output chimeric SAM file?
>> 
>> David Hoover
>> 
>> 
>> On 9/24/2014 7:03 PM, Ross wrote:
>>> Hi All,
>>> 
>>> That (fubar in testtoolshed) star wrapper was derived from one originally 
>>> written by Jeremy Goecks. I modified it for multiple inputs and added a few 
>>> tweaks and it has been in production use in our group for about 6 months so 
>>> I'm pretty sure it works reasonably well in our hands at least. 
>>> 
>>> I would really appreciate any available help getting it to a proven useful 
>>> state - suggestions and code welcomed. I have not moved it to the main 
>>> toolshed because aside from some encouragement, I've had no feedback to 
>>> suggest it's working - or not. It is extremely fast - we regularly see 
>>> 200-300M reads per minute in the logs!
>>> 
>>> We regularly run a whole experiment worth (eg 12 - 24) fastq files 
>>> simultaneously with the shared memory option working on our cluster - see 
>>> the readme.
>>> 
>>> Star index files made with a gene model (requires valid gff3) are huge - 
>>> 20-30GB for hg19 - hence the need for shared memory if you run multiple 
>>> jobs. That will eventually become a serious problem if you really want to 
>>> allow users to make their own - we definitely do not. You need to be very 
>>> careful about matching the gene model gff3 file to the reference and I had 
>>> enough trouble getting it right for the few major genomes we use to make me 
>>> think that I do not want users trying to do that generating 25GB of rubbish 
>>> every time they get it wrong.
>>> 
>>> There are challenges to do with needing different indexes for different 
>>> length reads but we are seeing fairly consistent 60bp single ended reads 
>>> for most of the incoming RNA seq experiments.
>>> 
>>> A data manager would be a boon if anyone cares to write one...
>>> 
>>> 
>>> On Thu, Sep 25, 2014 at 6:55 AM, Curtis Hendrickson (Campus) 
>>> <curt...@uab.edu> wrote:
>>> Bjorn
>>> 
>>> We'd be interested in this tool, as well. Any idea how close to functional 
>>> it is?
>>> I see it's only on TEST toolshed, and not on production, at this point.
>>> 
>>> I don't see any related Trello card when searching on "star"
>>> 
>>> Regards,
>>> Curtis
>>> Galaxy Admin @ University of Alabama at Birmingham
>>> 
>>> -----Original Message-----
>>> From: galaxy-dev-boun...@lists.bx.psu.edu 
>>> [mailto:galaxy-dev-boun...@lists.bx.psu.edu] On Behalf Of Björn Grüning
>>> Sent: Wednesday, September 24, 2014 3:15 PM
>>> To: galaxy-dev@lists.bx.psu.edu; hoove...@helix.nih.gov >> David Hoover
>>> Subject: Re: [galaxy-dev] tool for STAR RNA-seq aligner
>>> 
>>> Hi David,
>>> 
>>> yes there is inital code in the https://testtoolshed.g2.bx.psu.edu/. I 
>>> think Ross has done some work on it.
>>> The main problem with Star is that is needs special indices (and a lot of 
>>> it) and it would be great to offer data managers for it.
>>> 
>>> Cheers,
>>> Bjoern
>>> 
>>> Am 24.09.2014 um 22:05 schrieb David Hoover:
>>>> Hi,
>>>> 
>>>> I am developing a tool for STAR (https://code.google.com/p/rna-star/), and 
>>>> I realize I may be reinventing another wheel.  Has anyone else created a 
>>>> tool for STAR?  There's nothing else in the toolsheds for it yet.
>>>> 
>>>> David
>>>> 
>>>> --------------------
>>>> David Hoover, PhD
>>>> Helix Systems Staff
>>>> SCB/DCSS/CIT/NIH
>>>> 301-435-2986
>>>> http://helix.nih.gov
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> ___________________________________________________________
>>>> Please keep all replies on the list by using "reply all"
>>>> in your mail client.  To manage your subscriptions to this and other
>>>> Galaxy lists, please use the interface at:
>>>>   http://lists.bx.psu.edu/
>>>> 
>>>> To search Galaxy mailing lists use the unified search at:
>>>>   http://galaxyproject.org/search/mailinglists/
>>>> 
>>> ___________________________________________________________
>>> Please keep all replies on the list by using "reply all"
>>> in your mail client.  To manage your subscriptions to this and other Galaxy 
>>> lists, please use the interface at:
>>>  http://lists.bx.psu.edu/
>>> 
>>> To search Galaxy mailing lists use the unified search at:
>>>  http://galaxyproject.org/search/mailinglists/
>>> 
>>> ___________________________________________________________
>>> Please keep all replies on the list by using "reply all"
>>> in your mail client.  To manage your subscriptions to this
>>> and other Galaxy lists, please use the interface at:
>>>  http://lists.bx.psu.edu/
>>> 
>>> To search Galaxy mailing lists use the unified search at:
>>>  http://galaxyproject.org/search/mailinglists/
>>> 
>> 
>> 
> 
> 
> ___________________________________________________________
> Please keep all replies on the list by using "reply all"
> in your mail client.  To manage your subscriptions to this
> and other Galaxy lists, please use the interface at:
>  http://lists.bx.psu.edu/
> 
> To search Galaxy mailing lists use the unified search at:
>  http://galaxyproject.org/search/mailinglists/


___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client.  To manage your subscriptions to this
and other Galaxy lists, please use the interface at:
  http://lists.bx.psu.edu/

To search Galaxy mailing lists use the unified search at:
  http://galaxyproject.org/search/mailinglists/

Reply via email to